universal vote-by-mail has no impact on partisan turnout or vote … · 2020-05-07 · the e ect of...
TRANSCRIPT
Universal Vote-by-Mail Has No Impact on Partisan
Turnout or Vote Share
Daniel M. Thompson∗
Jennifer Wu†
Jesse Yoder,‡
Andrew B. Hall§
Democracy & Polarization Lab, Stanford University
May 6, 2020
Classification: SOCIAL SCIENCES: Political Sciences
∗Ph.D. Candidate, Department of Political Science; 616 Jane Stanford Way, Stanford, CA [email protected].†Pre-Doctoral Fellow, Stanford Institute of Economic Policy Research; 616 Jane Stanford Way, Stanford,
CA 94305. [email protected].‡Ph.D. Candidate, Department of Political Science; 616 Jane Stanford Way, Stanford, CA 94305. yo-
[email protected].§Corresponding author. Professor, Department of Political Science. Senior Fellow, Stanford Institute
of Economic Policy Research. 616 Jane Stanford Way, Stanford, CA 94305; (650)-723-1806. [email protected].
1
Abstract: In response to COVID-19, many scholars and policymakers are urging the U.S. to
expand voting-by-mail programs to safeguard the electoral process. What are the effects of
vote-by-mail? In this paper, we provide a comprehensive design-based analysis of the effect
of universal vote-by-mail—a policy under which every voter is mailed a ballot in advance of
the election—on electoral outcomes. We collect data from 1996-2018 on all three U.S. states
who implemented universal vote-by-mail in a staggered fashion across counties, allowing us
to use a difference-in-differences design at the county level to estimate causal effects. We find
that: (1) universal vote-by-mail does not appear to affect either party’s share of turnout;
(2) universal vote-by-mail does not appear to increase either party’s vote share; and (3) uni-
versal vote-by-mail modestly increases overall average turnout rates, in line with previous
estimates. All three conclusions support the conventional wisdom of election administration
experts and contradict many popular claims in the media.
Keywords: Vote-by-Mail; Elections; COVID-19; partisanship
Significance Statement: In response to COVID-19, many scholars and policymakers are
urging the U.S. to expand voting-by-mail programs to safeguard the electoral process, but
there are concerns that such a policy could favor one party over the other. We estimate
the effects of universal vote-by-mail, a policy under which every voter is mailed a ballot in
advance of the election, on partisan election outcomes. We find that universal vote-by-mail
does not affect either party’s share of turnout or either party’s vote share. These conclusions
support the conventional wisdom of election administration experts and contradict many
popular claims in the media. While it is difficult to predict whether either party might
benefit from the expansion of vote-by-mail during an unprecedented public health crisis, our
results imply that the outcomes of vote-by-mail elections in normal times closely resemble
in-person elections.
2
1 Introduction
The COVID-19 pandemic threatens the 2020 U.S. election. Fears that the pandemic could
deter many people from voting—or cause them to become infected if they do vote—have
spurred calls for major electoral reforms. As election administration experts Nathaniel
Persily and Charles Stewart put it: “The nation must act now to ensure that there will
be no doubt, regardless of the spread of infection, that the elections will be conducted on
schedule and that they will be free and fair.”1
Persily and Stewart recommend expanding vote-by-mail programs to allow Americans
the opportunity to vote from the safety from their own homes, but many question the
potential political consequences of such a policy. President Trump declared that, if it was
implemented, “you’d never have a Republican elected in this country again.”2 On the other
hand, Brian Dunn, a former Obama campaigner and founder of a company that works
on vote-by-mail programs, says that “There is justified concern that Democratic-leaning
voters may be disadvantaged through vote-by-mail systems.”3 This debate continues in part
because, in the academic literature, as Charles Stewart points out, “evidence so far on which
party benefits [has] been inconclusive.”4
We expand the existing evidence on the partisan effects of vote-by-mail programs by col-
lecting new data on voting and election outcomes in California and Utah, which we combine
with data on Washington state originally from Gerber, Huber, and Hill (2013), extended to
present day in our study. Together, this dataset allows us to study the full universe of county-
level universal vote-by-mail programs with staggered roll-outs. Universal vote-by-mail is the
strongest form of vote-by-mail; in all three states we study, every registered voter is sent a
ballot and in-person voting options decrease dramatically. Policy experts and policymakers
are primarily recommending that states without robust pre-existing vote-by-mail programs
1https://www.lawfareblog.com/ten-recommendations-ensure-healthy-and-trustworthy-2020-
election2https://www.theguardian.com/us-news/2020/apr/08/trump-mail-in-voting-2020-election3https://www.nytimes.com/2020/04/10/us/politics/vote-by-mail.html4https://www.nytimes.com/2020/04/10/us/politics/vote-by-mail.html
3
expand access by lifting requirements that voters provide a valid excuse in order to receive
an absentee ballot, while stopping short of moving to universal vote-by-mail5—as such, by
studying a more dramatic version of the recommended policies, our paper provides a useful
upper bound related to these discussions. While a large literature in political science studies
various forms of convenience voting—see Table S1 for a full review—there has not been any
comprehensive analysis of vote-by-mail that employs clear designs for causal inference to
estimate effects on partisan outcomes.6 The existing research supporting the neutral parti-
san effects of vote-by-mail compares turnout in Oregon before and after it implemented its
statewide universal vote-by-mail reform, or extrapolates from the behavior of irregular voters
to make predictions about partisan effects (Karp and Banducci 2000; Berinsky, Burns, and
Traugott 2001; Berinsky 2005).7
We find that implementing universal vote-by-mail has no apparent effect on either the
share of turned-out voters who are Democrats or the share of votes that go to Democratic
candidates, on average, although these latter estimates are a bit less precise. We also find
5For example, the proposed Klobuchar-Wyden bill calls for all Americans to have access to no-excuseabsentee voting. See https://www.klobuchar.senate.gov/public/_cache/files/0/0/00bfcd4c-8bff-
4e40-8082-9c509e6bf168/C874798F600EED86B58DED75D3FE5875.naturaldisasterandeba.pdf.6The existing papers with clear causal designs for the effect of universal vote-by-mail study overall turnout(Gerber, Huber, and Hill 2013), roll-off (Marble 2017), political information and accountability (Szewczyk2020a,b), the participation of low-propensity voters (Gerber, Huber, and Hill 2013), or precinct-level ratherthan county-level interventions (Elul, Freeder, and Grumbach 2017), and only study one state at a time.The closest analogue to the effect of universal vote-by-mail on a party’s vote share comes from Gerber,Huber, and Hill (2013) which finds that the turnout rates of high-propensity voters increase by less thanthose of low-propensity voters, who some may assume have different political leanings from regular voters.Fowler (N.d.) explicitly estimates the heterogeneous turnout effects of a convenience voting reform in WestVirginia using a county-level difference-in-difference design and finds no evidence for different effects byparty. Yet, the logic that expanding the pool of voters may favor one party is not flawed–for example,compulsory voting laws appear to improve the performance of the Labor party in Australia (Fowler 2013).Table S1 in the Appendix summarizes the large existing literature on vote-by-mail reforms which generallystudies the effect on turnout with findings ranging from a large increase Magleby (1987); Southwell andBurchett (2000); Richey (2008); Larocca and Klemanski (2011), to a modest increase or null effect (Berinsky,Burns, and Traugott 2001; Gronke, Galanes-Rosenbaum, and Miller 2007; Southwell 2009; Gerber, Huber,and Hill 2013; Menger, Stein, and Vonnahme 2015), to a decrease (Kousser and Mullin 2007; Elul, Freeder,and Grumbach 2017).
7Meredith and Malhotra (2011) presents evidence that vote-by-mail can change primary election outcomessince many voters mail their ballots before candidates withdraw. Szewczyk (2020b) finds the introductionof vote-by-mail in Washington lead to less taxation.
4
that it increases turnout by roughly 2 percentage points, on average—very similar to the
estimate reported in Gerber, Huber, and Hill (2013) for Washington state.
These findings are consistent with the conventional wisdom in the convenience-voting lit-
erature (see Gronke et al. (2008) for a review). However, they should increase our confidence
in these views, both because our data permits a stronger research design than was previously
possible and because our dataset runs through the 2018 midterm elections, allowing for the
most up-to-date analysis available.
Three main caveats are warranted in interpreting our findings. First, our evidence is
about the effects of counties opting into universal vote-by-mail programs during normal
times—that is, the counterfactual we are comparing voting-by-mail to is a normally admin-
istered in-person election. The effect of vote-by-mail programs relative to the counterfactual
of an in-person election during COVID-19 might be quite different, and the effect would
depend on whether we believe COVID-19 disproportionately deters Democrats or Republi-
cans from voting. In addition to being unsure what the effects of expansions of vote-by-mail
might be in the context of COVID-19, we should also stress that our focus is on the causal
effects of implementing vote-by-mail programs, and not on raw correlations between those
states that expand vote-by-mail and partisan outcomes. As the issue of vote-by-mail becomes
increasingly partisan, it is possible that Democratic-leaning states will lean into expanding
vote-by-mail more than Republican states. If this occurs, the subsequent correlation between
vote-by-mail expansions and the Democratic leanings of the electorate in these states will
not necessarily indicate that vote-by-mail caused these states to become more Democratic.
Second, our results say nothing about whether vote-by-mail should be implemented na-
tionwide. There may be reasons to worry about rolling out nationwide vote-by-mail that
we cannot study; for example, it might have disparate impact on minority voters, who some
claim utilize vote-by-mail at a lower rate (though also see McGhee et al. (2019)), 8 or it may
8https://www.nytimes.com/2020/04/10/us/politics/vote-by-mail.html
5
simply be too expensive to administer to be worth the cost. Finally, even if vote-by-mail did
have partisan effects, there might still be good reasons to support it as a policy.
Third, and finally, our paper directly studies the effects of what we call universal vote-by-
mail programs—the policy in which states mail every single registered voter a ballot. Many
of the policy proposals for the 2020 election fall short of universal vote-by-mail, and instead
focus on expanding opportunities for voters to opt into voting absentee. We do not have
direct evidence on the effects of these “no-excuse” vote-by-mail programs, but the universal
vote-by-mail programs we study represent a more dramatic intervention than no-excuse vote-
by-mail. We suspect, therefore, that univeral vote-by-mail might provide an upper bound
on the effect of no-excuse vote-by-mail.
Even with these caveats, our paper has a clear takeaway: claims that vote-by-mail fun-
damentally advantages one party over the other appear overblown. In normal times, based
on our data at least, vote-by-mail modestly increases participation while not advantaging
either party.
2 Voting-by-Mail and County Roll Outs
Led by Oregon in 2000, six states in the US have now adopted, or are in the process of
adopting, universal vote-by-mail elections.9 In some of these cases, the state has implemented
the vote-by-mail program across the entire state. For example, Oregon, Colorado, and
Hawaii made statewide switches to vote-by-mail elections beginning in 2000, 2014, and 2020,
respectively.10 Estimating the effects of these statewide adoptions of universal vote-by-mail
on partisan election outcomes, turnout, and the partisan composition of the electorate is
difficult, as these switches happen concurrently with other statewide changes and provide
no within-state counterfactuals.
9Colorado, Hawaii, Oregon, Utah, and Washington now conduct all elections by mail (see https://www.
ncsl.org/research/elections-and-campaigns/all-mail-elections.aspx).10We summarize these changes in Table S2 in the Appendix.
6
To study the effect of switching to universal vote-by-mail elections, in which all registered
voters are sent a ballot and nearly all votes are cast by mail, we narrow our focus to the
three states that rolled out universal vote-by-mail at the county-level in a staggered fashion:
California, Utah, and Washington.11 By comparing counties that adopt a vote-by-mail
program to counties within the same state that do not adopt the program, we are able
to compare the election outcomes and turnout behavior of voters who have different vote-
by-mail accessibility but who have the same set of candidates on the ballot for statewide
races.
Each of these three states’ reforms are slightly different, but all share a similar feature:
counties adopting universal vote-by-mail mailed an absentee ballot to every eligible voter in
the county, not just voters who had requested receiving a mailed absentee ballot. Voters
can mail their completed ballot to their county elections office, or deposit their ballot in
secure ballot drop-off locations throughout the county. Alternatively, each of these states’
reforms also replaces traditional polling places with some form of in-person voting, though
these options vary considerably by state.12
In Utah and Washington, each county has now adopted the vote-by-mail program de-
scribed above. In California, the county-level roll-out is ongoing. Following the adoption of
11In California, Utah, and Washington, vote-by-mail has become increasingly common. Figure 1 shows theshare votes cast in the general election that are vote-by-mail, in California and Washington in each electionyear from 1998-2018. In the late 1990s, the majority of votes cast in both states came from non-VBMoptions. By the late 2010s, nearly every county in California had a majority of their votes cast using VBMand Washington had all-mail elections.
12In California, counties that adopt all-mail elections are required to have one in-person voting centerfor every 10,000 registered voters on election day (see https://www.sos.ca.gov/elections/voters-
choice-act/about-vca/). Utah also offers some opportunities for in-person voting in existing gov-ernment offices to ensure those with disabilities or issues with their ballots are able to participate (seehttps://www.youtube.com/watch?v=LkZntyrpNyQ). As of 2011, all counties in Washington were requiredto have at least one in-person voting center for general, primary, and special elections (see “voting cen-ters” at https://app.leg.wa.gov/RCW/default.aspx?cite=29a.40). At least some vote-by-mail coun-ties in Washington had an in-person voting option prior to 2011 to comply with the federal Help AmericaVote Act (see, e.g., https://web.archive.org/web/20061031134553/http://www.kitsapgov.com/aud/elections/disabilityaccess.htm).
7
Figure 1 – Increase in the Adoption of Universal Vote-by-Mail,California, Utah, and Washington General Elections, 1996 to 2018.
Washington
Utah
California
0
.2
.4
.6
.8
1Sh
are
of C
ount
ies
w A
ll-M
ail E
lect
ions
1996 2000 2004 2008 2012 2016 2020Year
California’s Voter’s Choice Act (VCA), 5 of California’s 58 counties adopted universal vote-
by-mail for the 2018 elections, followed by an additional 10 counties for the 2020 elections.13
Figure 1 shows the timing of each state’s county-level roll out of vote-by-mail reforms,
and it illustrates the main source of variation we exploit in this study. The vertical axis
represents the share of counties in each of the three states we study that adopt universal
vote-by-mail. As we can see, each state rolled out its vote-by-mail program in a staggered
fashion over several election cycles.
2.1 Outcomes of Interest
We collect data on a variety of outcomes to see how universal vote-by-mail might affect
elections. First, we are interested in how vote-by-mail affects the performance of Democratic
versus Republican candidates. We collect county-level general election results for each state
13For the 2018 election 14 of California’s 58 counties were allowed to opt into this new format for conductingelections, and all of California’s counties were allowed to adopt these changes beginning in 2020. Seehttps://www.sos.ca.gov/elections/voters-choice-act/about-vca/.
8
from its Secretary of State website to construct the Democratic two-party vote share in each
Presidential, Gubernatorial, and Senatorial general election.14
Second, we are interested in how vote-by-mail might affect the partisan composition
of the electorate. For this outcome, we use the California and Utah voter files, provided
by L2, a private data vendor. The voter files contain information on each individual’s
name, registration address, date of birth, date of registration, party registration, and turnout
history. Using the voter file, we can observe whether universal vote-by-mail led to a more
Democratic or Republican electorate, based on the party registration of those who turn out
to vote.
Finally, we are interested in the effect of vote-by-mail on turnout and vote-by-mail usage.
For California and Utah, we collect the number of ballots cast in each general election from
official state sources. For Washington, we use turnout provided by Gerber, Huber, and Hill
(2013). To construct a turnout share, we divide the total number of ballots cast by the
county’s citizen voting age population in that year.15 For California and Washington, we
also observe each county’s turnout by vote mode, so we can construct a measure of the share
of total votes in a county that come via vote-by-mail.
Table 1 summarizes the information that we have collected from each of the three states
that we study. Overall, the data we have collected covers a wide range of years (1996-2018).
It includes each election cycle’s turnout and election results from all three states. Vote-
by-mail usage comes from California and Washington, and our analyses on the partisan
composition of the electorate that use the voter file come from California.
14In California, we use only Gubernatorial, not Presidential, election results. This is because the earliestcounty California to adopt universal vote-by-mail was in 2018, and a Presidential general election has notyet occurred since then.
15Each county’s citizen voting age population is collected from https://www.census.gov/programs-
surveys/decennial-census/about/voting-rights/cvap.html.
9
Table 1 – Information Included in Various Data Sources. Each columndenotes a state, and checkmarks indicate features or observable information inthat state. Turnout data is missing in California for the year 2000. While wehave presidential election data for California, it did not implement its vote-by-mail program until after the 2016 presidential election. Similarly, while we havesenatorial election data for California, it implemented a top-two primary systemand its general election race for Senate included two Democrats in 2018.
California Utah Washington
General Election Turnout " " "
Vote-by-Mail Ballot Usage " "
Gubernatorial Election Results " " "
Senatorial Election Results N/A " "
Presidential Election Results N/A " "
Voter File " "Years Included 1998-2018 1996-2018 1996-2016
3 Empirical Approach: Difference-in-Differences
Estimating the effect of vote-by-mail programs is difficult because the states that have im-
plemented vote-by-mail differ systematically from those that have not. Figure S1 in the
Appendix shows that states that go on to adopt universal VBM (those listed in Table S2)
are states that have had higher average Democratic vote shares for President, on average,
than states that do not adopt these policies. Moreover, the gap in Democratic vote shares in
VBM states an non-VBM states has grown over time. If we found, for example, that vote-
by-mail programs are correlated with higher turnout for Democratic voters using a statewide
design, we could not conclude that vote-by-mail causes Democratic voters to turn out more;
it could be that Democratic voters simply turn out to vote more in liberal states. To get
at the actual effect of the vote-by-mail program, we need to approximate an experiment in
which some elections occur under vote-by-mail while other, similar elections do not.
To do something like this, we take advantage of the staggered rolling out of vote-by-
mail across counties, within California, Utah, and Washington, as we explained above. In
10
particular, we estimate the following equation:
Ycst = βVBMcst + γcs + δst + εcst, (1)
where Y is an outcome variable—usually partisan turnout rates or vote share—in county
c in state s during election t. Our treatment indicator, V BM , takes a value of 1 if the
county opts into its state’s vote-by-mail program, and 0 otherwise. The γcs and δst terms
represent county fixed effects and state-by-election fixed effects, respectively. As the above
equation makes clear, this is a difference-in-differences design, where we compare within-
county changes in turnout over time across changes in vote-by-mail policy. To identify β as
the causal effect of universal vote-by-mail, it must be the case that the trends in turnout
in counties that do not adopt vote-by-mail provide valid counterfactuals for the trends we
would have observed in the treatment counties, had they chosen not to adopt vote-by-mail.
We use a variety of tests to evaluate whether the parallel trends assumption might be
reasonable in our case. First, to test for anticipatory effects, following Angrist and Pischke
(2009) we plot coefficients on leads of our outcome variables and compare them to our
estimated treatment effects. The simple idea of these tests is that a county’s vote-by-mail
program should not affect our outcomes in the elections prior to its adoption. Second, we
relax the parallel trends assumption in a variety of ways by including more flexible sets of
fixed effects, like linear or quadratic time trends. We discuss these tests in detail throughout
the results sections.
11
4 Neutral Partisan Effects of Vote-by-Mail
Does vote-by-mail favor either political party in elections? Table 2 presents our main re-
sults.16 The first column shows our basic state-specific difference-in-differences design where
the outcome is the share of voters—that is, people who turn out to vote—who are Democrats.
In this specification, we estimate that the Democratic turnout share increases by 0.7 per-
centage points as a result of vote-by-mail. This specification uses state-by-year fixed effects,
estimating state-specific time shocks, and therefore makes the within-state comparisons that
Gerber, Huber, and Hill (2013) recommends.
In columns 2 and 3, we also examine the possibility that counties may be on different
trends by including linear (column 2) and quadratic (column 3) county-specific time trends.
The inclusion of these trends attenuates the estimates dramatically, to only 0.1 percentage
points, while also shrinking the standard errors. In the latter two specifications, which are
our most precise specifications, even the upper bound of the 95% confidence interval is only
about 0.3 percentage points, a very small effect. We conclude from these estimates that,
while our simplest difference-in-differences estimate suggests a small but detectable effect on
Democratic share of turnout, more plausible estimates suggest a truly negligible effect.
The latter three columns use the same set of specifications to explore the difference-in-
differences estimates for the effect of vote-by-mail on Democratic candidate two-party vote
share, pooling together Democratic gubernatorial candidates, Democratic senate candidates,
and Democratic presidential candidates.17 In column 4, when we use state-by-year fixed
effects without time trends, we estimate a 2.8 percentage-point increase for Democrats—
however, when we add trends in columns 5 and 6, this estimate attenuates markedly. While
16The in-person voting options vary some by state, as we discussed in Section 2. For this reason, andsince these states vary in other ways, we show the results separately for each state in Section S8 of theAppendix. The results are reassuring. In particular, we do not see any evidence of a larger effect ofvote-by-mail expansion in Washington, the state with the most extreme expansion. The estimates appearto be similarly null in all three contexts.
17The number of counties increases in columns 4-6 because we have data from all three states, whereas incolumns 1-3 we have partisan turnout data from California and Utah. In Table S3 in the Appendix, weshow the same version of Table 2, but using only California and Utah for all six columns. The resultsremain substantively similar.
12
Table 2 – Vote-by-Mail Expansion Does Not Appear to Favor Ei-ther Party.
Dem Turnout Share [0-1] Dem Vote Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.007 0.001 0.001 0.028 0.011 0.007(0.003) (0.001) (0.001) (0.011) (0.004) (0.003)
# Counties 87 87 87 126 126 126# Elections 23 23 23 31 31 31# Obs 986 986 986 1,998 1,998 1,998
County FE Yes Yes Yes Yes Yes YesState by Year FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Robust standard errors clustered by county in parentheses. The number of counties issmaller in columns 1-3 because we have partisan turnout share for CA and UT, butnot WA. Columns 4-6 use data from all three states.
the standard errors on these estimates are larger than the standard errors on the turnout
share estimates, they continue to suggest modest or null effects, and they are nowhere near
the magnitude necessary to represent a major, permanent electoral shift towards the Demo-
cratic party.18
In sum, looking across turnout and vote share outcomes, the substantively small size
of the estimated effects leads us to conclude that vote-by-mail does not have meaningful
partisan effects on election outcomes. We find the estimates on the Democratic share of
turnout, which are particularly precise, to be most compelling. Universal vote-by-mail does
not appear to tilt turnout towards the Democratic party, nor does it appear to affect election
outcomes meaningfully.
18We show graphical evidence of the neutral partisan effects of vote-by-mail in Figure S5 in the Appendix.
13
5 Universal Vote-by-Mail Modestly Increase Turnout
Having evaluated the partisan effects of vote-by-mail, we now evaluate its effect on political
participation as measured by the share of the eligible population that turns out to vote in
general elections.
Table 3 presents formal estimates of the effect of universal vote-by-mail on participation.19
The first three columns report estimates of the effect on the number of voters participating
as a share of the citizen voting-age population. As in Table 2, Column 1 reports reports the
within-state estimate, and columns 2 and 3 add linear and quadratic county-specific trends,
respectively. Looking across the columns, we see a stable estimate showing that vote-by-
mail causes around a 2-percentage-point increase (estimates range from 2.1 to 2.2 percentage
points) in the share of the citizen voting-age population that turns out to vote.20
The final three columns, using the same regression specifications as columns 1 through
3, show that universal vote-by-mail produces a large increase in the share of ballots that are
mailed in—roughly a 14 to 19 percentage-point increase across specifications. This is not a
surprising finding, but it does show that large numbers of voters appreciate the chance to
mail in their ballot.21
6 Conclusion
This paper has offered new data to offer the most up-to-date, most credible causal evidence
on the effects of universal vote-by-mail programs on partisan electoral outcomes and partici-
pation during normal times. In our data, we confirm important conventional wisdom among
election experts: vote-by-mail offers voters considerable convenience, increases turnout rates
modestly, but has no discernible effect on party vote shares or the partisan share of the
electorate.
19We show the results separately for each state in Section S9 of the Appendix.20We show graphical evidence of the participation effect in Figure S6 in the Appendix.21Existing work on universal vote-by-mail in California and Oregon reaches a similar conclusion that voters
take advantage of the opportunity to vote by mail (Southwell 2004; Bryant 2019).
14
Table 3 – Vote-by-Mail Expansion Increases Participation.
Turnout Share [0-1] Vote-by-Mail Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.021 0.022 0.021 0.186 0.157 0.136(0.009) (0.007) (0.008) (0.027) (0.035) (0.085)
# Counties 126 126 126 58 58 58# Elections 30 30 30 10 10 10# Obs 1,240 1,240 1,240 580 580 580
County FE Yes Yes Yes Yes Yes YesState by Year FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Robust standard errors clustered by county in parentheses.
Our results should strengthen the field’s confidence in these effects of vote-by-mail. While
the design we implement is by no means perfect, our new data does permit empirical ap-
proaches stronger than those used in the existing literature. Only one existing paper in
the vote-by-mail literature employs a similar design, and it studies only participation and
only in the state of Washington. As such, we believe our paper is the most comprehensive
confirmation to date of vote-by-mail’s neutral partisan effects.
As the country debates how to run the 2020 election in the shadow of COVID-19, politi-
cians, journalists, pundits, and citizens will continue to hypothesize about the possible effects
of vote-by-mail programs on partisan electoral fortunes and participation. We hope that our
study will provide a useful data point for these conversations.
15
Acknowledgements
For helpful comments and suggestions, the authors thank Alex Coppock, Bernard Fraga,
Paul Gronke, Greg Huber, Alisa Hall, Thad Kousser, Shiro Kuriwaki, Eric McGhee, Nate
Persily, and Charles Stewart, as well as Utah’s Director of Elections Justin Lee, Amelia
Showalter from Pantheon Analytics, and county elections offices in California and Utah.
The authors also thank Alan Gerber, Greg Huber, and Seth Hill for generously sharing data.
16
References
Angrist, Joshua D, and Jorn-Steffen Pischke. 2009. Mostly Harmless Econometrics: AnEmpiricist’s Companion. Princeton University Press.
Berinsky, Adam J. 2005. “The Perverse Consequences of Electoral Reform in the UnitedStates.” American Politics Research 33(4): 471–491.
Berinsky, Adam J, Nancy Burns, and Michael W Traugott. 2001. “Who Votes by Mail? ADynamic Model of the Individual-Level Consequences of Voting-By-Mail Systems.” PublicOpinion Quarterly 65(2): 178–197.
Bryant, Lisa A. 2019. “Changing the Way People Vote? An Examination of the VoterChoice Act and Vote Center Implementation in California.” Election Sciences, Reformand Administration Conference. Working Paper. https://cpb-us-w2.wpmucdn.com/web.sas.upenn.edu/dist/7/538/files/2019/07/ESRA2019_Bryant.pdf.
Elul, Gabrielle, Sean Freeder, and Jacob M Grumbach. 2017. “The Effect of Mandatory MailBallot Elections in California.” Election Law Journal: Rules, Politics, and Policy 16(3):397–415.
Fowler, Anthony. 2013. “Electoral and Policy Consequences of Voter Turnout: Evidence fromCompulsory Voting in Australia.” Quarterly Journal of Political Science 8(2): 159–182.
Fowler, Anthony. N.d. “Promises and Perils of Mobile Voting.” Forthcoming, Election LawJournal .
Gerber, Alan S., Gregory A. Huber, and Seth J. Hill. 2013. “Identifying the Effect of All-Mail Elections on Turnout: Staggered Reform in the Evergreen State.” Political ScienceResearch and Methods 1(1): 91–116.
Gronke, Paul, Eva Galanes-Rosenbaum, and Peter A Miller. 2007. “Early Voting andTurnout.” PS: Political Science & Politics 40(4): 639–645.
Gronke, Paul, Eva Galanes-Rosenbaum, Peter A Miller, and Daniel Toffey. 2008. “Conve-nience Voting.” Annual Review of Political Science 11: 437–455.
Karp, Jeffrey A, and Susan A Banducci. 2000. “Going Postal: How All-Mail ElectionsInfluence Turnout.” Political Behavior 22(3): 223–239.
Kousser, Thad, and Megan Mullin. 2007. “Does Voting by Mail Increase Participation?Using Matching to Analyze a Natural Experiment.” Political Analysis 15(4): 428–445.
Larocca, Roger, and John S Klemanski. 2011. “US State Election Reform and Turnout inPresidential Elections.” State Politics & Policy Quarterly 11(1): 76–101.
Magleby, David B. 1987. “Participation in Mail Ballot Elections.” Western Political Quar-terly 40(1): 79–91.
17
Marble, William. 2017. “Mail Voting Reduces Ballot Roll-Off: Evidence from WashingtonState.” Working Paper. https://williammarble.co/docs/rolloff_vbm.pdf.
McGhee, Eric, Mindy Romero, Laura Daly, and Thad Kousser. 2019. “New Elec-torate Study: How Did the Voter’s Choice Act Affect Turnout in 2018?” Re-search Brief. https://uccs.ucdavis.edu/events/event-files-and-images/
ResearchBriefHowDidtheVCAAffectTurnoutin2018FINAL2.pdf.
Menger, Andrew, Robert M Stein, and Greg Vonnahme. 2015. “Turnout Effects fromVote by Mail Elections.” Conference on Election Administration and Reform. WorkingPaper. https://static1.squarespace.com/static/599b81d8e6f2e1452a4955ea/t/
599bb45446c3c42867a1a535/1503376469774/Turnout+Effects+from+VBM+Elections.
pdf.
Meredith, Marc, and Neil Malhotra. 2011. “Convenience Voting Can Affect Election Out-comes.” Election Law Journal 10(3): 227–253.
Richey, Sean. 2008. “Voting by Mail: Turnout and Institutional Reform in Oregon.” SocialScience Quarterly 89(4): 902–915.
Southwell, Priscilla L. 2004. “Five Years Later: A Re-Assessment of Oregon’s Vote by MailElectoral Process.” Ps: Political Science & Politics 37(1): 89–93.
Southwell, Priscilla L. 2009. “Analysis of the Turnout Effects of Vote by Mail Elections,1980–2007.” The Social Science Journal 46(1): 211–217.
Southwell, Priscilla L, and Justin I Burchett. 2000. “The Effect of All-Mail Elections onVoter Turnout.” American Politics Quarterly 28(1): 72–79.
Szewczyk, James. 2020a. “Does Vote-by-Mail Cause Voters to Gather Information AboutPolitics?” Working Paper. http://jamesszewczyk.com/VBM_Information.pdf.
Szewczyk, James. 2020b. “How Electoral Institutions Affect Political Accountability: Ev-idence from All-Mail Elections.” Working Paper. http://jamesszewczyk.com/VBM_V7.pdf.
18
Supporting Information Appendix:
Universal Vote-by-Mail Has No Impact on Partisan Turnout
or Vote Share
Daniel M. Thompson, Jennifer Wu, Jesse Yoder, and Andrew B. Hall
Contents
S1 Summary of the Extant Literature on Vote-by-Mail Effects . . . . . . . . . . 20S2 Universal Vote-by-Mail Adoption by State . . . . . . . . . . . . . . . . . . . 21S3 Differences Between VBM and non-VBM States . . . . . . . . . . . . . . . . 22S4 Increasing Use of Vote-by-Mail . . . . . . . . . . . . . . . . . . . . . . . . . 23S5 Robustness of Composition Effects to Elections Included . . . . . . . . . . . 25S6 Graphical Evidence of Partisan Effects of Vote-by-Mail . . . . . . . . . . . . 26S7 Partisan Effects Limiting to California and Utah . . . . . . . . . . . . . . . 27S8 Partisan Effects by State . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28S9 Participation Effects by State . . . . . . . . . . . . . . . . . . . . . . . . . . 30S10 Graphical Evidence of Vote-by-Mail Effect on Participation . . . . . . . . . 32S11 Effect of Universal Vote-by-Mail on Republican Participation . . . . . . . . 33S12 Effects On Age of Electorate . . . . . . . . . . . . . . . . . . . . . . . . . . 35S13 Effects On Socio-Economic Status and Racial Composition of Electorate . . 37
19
S1 Summary of the Extant Literature on Vote-by-Mail
Effects
This section summarizes the literature to date on the effects of vote-by-mail programs. Eachrow of Table S1 represents a study on the effects of vote-by-mail, and the columns summarizethe study’s setting research design, effect on overall turnout, and a summary of its effect onthe composition of the electorate, if any.
Table S1 – Review of Vote-by-Mail Effects Literature. Note: Magelby(1987) studies a selection of cities in the United States and Canada. All othersettings are state abbreviations. X-Section refers to a cross-sectional design, andDiD refers to a difference-in-differences design.
Paper Setting Design Turnout Effect Composition Effect Partisan Effect
Magelby (1987) USA, CAN Pre-Post Large +Karp and Banducci (2000) OR Pre-Post Modest − to Modest + ↑ Frequent VotersSouthwell and Burchett (2000) OR Pre-Post Large +Berinsky, Burns, and Traugott (2001) OR Pre-Post Modest + ↑ Frequent Voters No EffectGronke et al. (2007) OR State Panel Modest +Kousser and Mullin (2007) CA X-Section Modest -Richey (2008) OR State Panel Modest/Large +Southwell (2009) OR Pre-Post Modest − to NullBergman and Yates (2011) CA Pre-Post Large −Larocca and Klemanski (2011) OR, WA X-Section Modest/Large +Gerber, Huber, and Hill (2013) WA County DiD Modest + ↑ Infrequent VotersMenger, Stein, and Vonnahme (2015) CO Pre-Post Modest +Elul et al. (2017) CA Precinct DiD Modest −Keele and Titiunik (2018) CA Geo RDD Modest −Atsusaka (2019) CO Pre-Post Modest + ↑ Infrequent Voters
20
S2 Universal Vote-by-Mail Adoption by State
In this section, we provide a summary of states that have adopted programs to mail everyregistered voter a ballot by default. Table S2 shows this information, along with the levelat which the vote-by-mail program was rolled out and some information about the timingof its implementation. As we can see, three states, Oregon, Colorado, and Hawaii haveimplemented statewide vote-by-mail programs in 2000, 2014, and 2020, respectively. Thethree states we focus on, California, Utah, and Washington, are those that have rolled outtheir vote-by-mail programs in a staggered fashion by county. The roll-out in California isstill ongoing.
Table S2 – States With Programs to Mail Every Registered Voter aBallot. Note: This table shows states where every county in the state is eligibleto adopt a program to mail every registered voter a ballot for primary and generalelections. Nebraska and North Dakota allow only some counties to conduct all-mailelections, and several other states allow some counties to conduct special electionsor local elections by mail.
State Level of Roll-Out Year Started Year FullyImplemented
California County 2018 OngoingColorado State 2014 2014Hawaii State 2020 2020Oregon State 2000 2020Utah County 2012 2020Washington County 2006 2010
21
S3 Differences Between VBM and non-VBM States
In this section, we show a key difference in the voting patterns of states that have adoptedvote-by-mail programs and those that have not. We collect state-level presidential electionresults for each state from 1992-2016. In Figure S1, we plot the Democratic Presidentialtwo-party vote share separately for states that adopt a VBM program at some point andthose that do not. The VBM states (those listed in Table S2) tend to vote for Democraticpresidential candidates at higher rates than non-VBM states. Moreover, this gap has in-creased over time: in recent presidential elections, the average Democratic presidential voteshare was about 10 percentage points higher in VBM states compared to non-VBM states.Overall, this illustrates the disadvantage of studying the effects of vote-by-mail programs atthe state-level. The six states that have adopted vote-by-mail programs not only tend tovote for Democratic candidates at higher rates, but they also are trending more quickly in aDemocratic direction than states that have not adopted VBM.
Figure S1 – Democratic Presidential Two-Party Vote Share overtime, by Vote-by-Mail. VBM states are those listed in Table S2.
Non-VBM States
VBM States
.45
.5
.55
.6
Dem
Vot
e Sh
are,
Pre
s
1992 1996 2000 2004 2008 2012 2016Year
22
S4 Increasing Use of Vote-by-Mail
In this section, we show the fraction of votes cast using vote-by-mail over time for Californiaand Washington. As we show, vote-by-mail usage has become increasingly common overtime in both states.
First, in Figure S2 we show vote-by-mail usage in California general elections over time.Each plot is a histogram of California counties, with the x-axis representing the share oftotal votes that were cast using vote-by-mail. As we can see, nearly all California countiesreceived less than half of their ballots from vote-by-mail in 1998, but by 2018 nearly allcounties in California received more than half of their ballots from vote-by-mail.
Figure S2 – Use of Vote-by-Mail in CA General Elections, 1998 to2018.
0
5
10
15
20
0
5
10
15
20
0
5
10
15
20
0 .5 1 0 .5 1
0 .5 1 0 .5 1
1998 2002 2004 2006
2008 2010 2012 2014
2016 2018
Num
of C
ount
ies
Mail Votes/Turnout
Next, we show the same set of histograms of vote-by-mail usage over time for Washington.Most of the counties adopted Washington’s switch to exclusively vote-by-mail starting in2006, which is where we see the largest shift toward vote-by-mail usage. By 2010, nearly allWashington counties had switched to the exclusive vote-by-mail program.
23
Figure S3 – Use of Vote-by-Mail in WA General Elections, 1996to 2010.
0
10
20
30
40
0
10
20
30
40
0
10
20
30
40
0 .5 1
0 .5 1 0 .5 1
1996 1998 2000
2002 2004 2006
2008 2010
Num
of C
ount
ies
Mail Votes/Turnout
24
S5 Robustness of Composition Effects to Elections In-
cluded
In this section, we show the robustness of our main results on the effects of vote-by-mail onthe partisan composition of the electorate (columns 1-4 of Table 2). For all of our resultson the composition of the electorate, we use the California voter file. One concern withusing this data is that voters removed from the voter file over time may be different fromthose remaining on the lists. In particular, we know that older voters in 1998 are much lesslikely to still be in the 2019 voter file we are using. This problem should be much smaller inelections that were held closer to the time when the voter file was compiled. In Figure S4,we evaluate the sensitivity of our results to the number of elections prior to 2018 that weinclude in the difference-in-differences regression. We find that the results are substantivelyunchanged when we include fewer elections and, if anything, shrink toward zero.
Figure S4 – Partisan Composition Effects Not Sensitive to YearsIncluded in Sample.
-.005
0
.005
.01
.015
Effe
ct o
n D
em S
hare
of E
lect
orat
e
1998 2002 2006 2010 2014First Year Included in Regression
25
S6 Graphical Evidence of Partisan Effects of Vote-by-
In this section, we provide a graphical examination of vote-by-mail’s effects on our outcomesin Table 2, which further suggests that there are no vote-by-mail effects on partisan turnoutor vote share beyond the pre-trending issue. Figure S5 plots estimated “effects” of vote-by-mail for three pre-treatment periods as well as for the actual treatment period. These areestimated by including four dummy variables in the regression corresponding to column 2in Table 2: three leads that take the value if the county became treated three elections inthe future, two in the future, or one in the future, as well as the standard treatment dummyindicating the the county was a vote-by-mail county. As the plot shows, the pre-treatmenteffects are nearly as large as the estimated post-treatment effect, and they trend upwardssteadily, with the estimated post-treatment effect essentially on trend. This further suggeststo us that even the small partisan vote-share effect we estimate in our regressions is likelyto be the result of residual pre-trending rather than a real effect.
Figure S5 – Vote-by-Mail Reform and Pre-Trends.
-.005
0
.005
.01
.015
.02
With
in-C
ount
y D
iff in
Dem
Tur
nout
Sha
re
-3 -2 -1 0Elections Since Treatment
0
.02
.04
.06
.08
.1
With
in-C
ount
y D
iff in
Dem
Vot
e Sh
are
-3 -2 -1 0Elections Since Treatment
26
S7 Partisan Effects Limiting to California and Utah
In this section, we show a version of Table 2 where we limit the analysis to Californiaand Utah. For columns 1-3, where the outcome is Democratic turnout share, the resultsremain the same as in Table 2 because we have partisan composition for those two states,but not for Washington. For the Democratic vote share results in columns 4-6, in the mainresults in our paper we pool all three states. Here, to make sure the sample we are studying isconsistent across the two outcome variables, we limit the vote share analysis to just Californiaand Utah. The results are slightly noisier because we have dropped Washington, but thetakeaway remains substantively similar to our main results in Table 2.
Table S3 – Vote-by-Mail Expansion Does Not Appear to FavorEither Party (California and Utah).
Dem Turnout Share [0-1] Dem Vote Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.007 0.001 0.001 0.041 0.009 0.004(0.003) (0.001) (0.001) (0.021) (0.007) (0.006)
# Counties 87 87 87 87 87 87# Elections 23 23 23 21 21 21# Obs 986 986 986 1,218 1,218 1,218
County FE Yes Yes Yes Yes Yes YesState by Year FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Robust standard errors clustered by county in parentheses.
27
S8 Partisan Effects by State
In this section, we show our effects of vote-by-mail on partisan outcomes separately foreach of the three states we study. The specifications in each of the table mirror those inour main results in Table 2. In each of these tables, we report confidence intervals forour estimates using the block bootstrap procedure described Cameron, Gelbach, and Miller(2008), clustered by county, due to the small number of clusters when estimating the effectsindependently for each state.
We show the results for California in Table S4. The effect on the Democratic turnoutshare is close to zero, and it is precisely estimated. The results for Democratic vote shareare also similar to when we pool across all three states. At first it appears that vote-by-mailmight lead to a small increase in Democratic vote share, but when we account for possiblepre-trending by including county trends, it becomes clear that the effect is close to zero.
Table S4 – Vote-by-Mail Expansion Does Not Appear to FavorEither Party in California.
Dem Turnout Share [0-1] Dem Vote Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.003 -0.004 -0.001 0.030 0.001 -0.012[-0.004,0.013] [-0.008,-0.001] [-0.009,0.005] [0.001,0.058] [-0.017,0.014] [-0.052,0.027]
# Counties 58 58 58 58 58 58# Elections 11 11 11 11 11 11# Obs 638 638 638 638 638 638
County FE Yes Yes Yes Yes Yes YesYear FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Block wild bootstrap confidence intervals clustered by county in brackets.
We show the results for Utah in Table S5. Similar to the results for California, it appearsat first that vote-by-mail might increase democratic vote shares by a small amount, butonce we move to our more plausible specifications in columns 5 and 6, it appears that theincreases we observe can be explained almost entirely by pre-trending.
Finally, we show the results on Democratic vote share for Washington in Table S6. Wedo not show the results for partisan share of turnout in Washington because we only havethat information for California and Utah. In each of the specifications in Table S6, the effectof vote-by-mail on the Democratic vote share is small.
28
Table S5 – Vote-by-Mail Expansion Does Not Appear to FavorEither Party in Utah.
Dem Turnout Share [0-1] Dem Vote Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.009 0.003 0.001 0.044 0.012 0.008[0.000,0.017] [-0.001,0.006] [-0.001,0.004] [-0.013,0.105] [-0.008,0.032] [-0.007,0.024]
# Counties 29 29 29 29 29 29# Elections 12 12 12 10 10 10# Obs 348 348 348 580 580 580
County FE Yes Yes Yes Yes Yes YesYear FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Block wild bootstrap confidence intervals clustered by county in brackets.
Table S6 – Vote-by-Mail Expansion Does Not Appear to FavorEither Party in Washington.
Dem Vote Share [0-1](1) (2) (3)
VBM 0.015 0.012 0.008[0.006,0.022] [0.004,0.020] [0.001,0.016]
# Counties 39 39 39# Elections 10 10 10# Obs 780 780 780
County FE Yes Yes YesYear FE Yes Yes YesCounty Trends No Linear Quad
Block wild bootstrap confidence intervals clustered bycounty in brackets.
29
S9 Participation Effects by State
In this section, we show our effects of vote-by-mail on participation outcomes separatelyfor each of the three states we study. The specifications in each of the table mirror thosein our main results in Table 3. In each of these tables, we report confidence intervals forour estimates using the block bootstrap procedure described Cameron, Gelbach, and Miller(2008), clustered by county, due to the small number of clusters when estimating the effectsindependently for each state.
We show the results for California in Table S7. We can see that vote-by-mail increasesturnout by about 1.4 to 1.8 percentage points in California, which is slightly smaller thanthe pooled effect we report in Table 3. Our results in columns 4-6 for the vote-by-mail shareare the same as Table 3 because we only use California in that analysis.
Table S7 – Vote-by-Mail Expansion Increases Participation in Cal-ifornia.
Turnout Share [0-1] Vote-by-Mail Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.018 0.014 0.014 0.186 0.157 0.136[-0.010,0.044] [-0.006,0.040] [-0.013,0.063] [0.123,0.259] [0.065,0.253] [-0.198,0.266]
# Counties 58 58 58 58 58 58# Elections 10 10 10 10 10 10# Obs 580 580 580 580 580 580
County FE Yes Yes Yes Yes Yes YesYear FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Block wild bootstrap confidence intervals clustered by county in brackets.
Next, we show the results for Utah in Table S5. In Utah, vote-by-mail appears to haveincreased turnout by a little over 3 percentage points, which is slightly higher than the pooledeffect we report in Table 3. We do not have information on vote-by-mail usage in Utah, sowe do not show results of vote-by-mail’s effect on the VBM share in Utah.
Finally, we show the results for Washington in Table S9. The effect of vote-by-mail onturnout in Washington hovers around 1 percentage point across specifications, which is lowerthan the pooled effect. Just looking at Washington alone, it looks like the effect of vote-by-mail on turnout is very modest. In columns 4-6, we show the effect of vote-by-mail onthe share of voters using VBM. Because the reform in Washington sends the vote-by-mailshare to 1 for all treated counties, the effect on the vote-by-mail share is massive. We do notinclude Washington in our main results because we wanted to measure voters’ preferencesfor vote-by-mail, given the option. For that reason, in the main results we subset just toCalifornia, where voters have the option to mail in their ballot or vote in person at a votingcenter in their county.
30
Table S8 – Vote-by-Mail Expansion Increases Participation inUtah.
Turnout Share [0-1](1) (2) (3)
VBM 0.032 0.032 0.034[-0.009,0.073] [0.006,0.058] [0.009,0.061]
# Counties 29 29 29# Elections 12 12 12# Obs 348 348 348
County FE Yes Yes YesYear FE Yes Yes YesCounty Trends No Linear Quad
Block wild bootstrap confidence intervals clustered bycounty in brackets.
Table S9 – Vote-by-Mail Expansion Increases Participation inWashington.
Turnout Share [0-1] Vote-by-Mail Share [0-1](1) (2) (3) (4) (5) (6)
VBM 0.009 0.011 0.004 0.300 0.312 0.304[-0.001,0.019] [-0.001,0.022] [-0.009,0.018] [0.226,0.367] [0.225,0.404] [0.175,0.436]
# Counties 39 39 39 39 39 39# Elections 8 8 8 8 8 8# Obs 312 312 312 312 312 312
County FE Yes Yes Yes Yes Yes YesYear FE Yes Yes Yes Yes Yes YesCounty Trends No Linear Quad No Linear Quad
Block wild bootstrap confidence intervals clustered by county in brackets.
31
S10 Graphical Evidence of Vote-by-Mail Effect on Par-
ticipation
Figure S6 presents visual evidence of the effect on turnout. Each point represent represents aregression coefficient with the first three points being leads that anticipate a county’s switchinto vote-by-mail by three elections, two elections, and one election. The fourth point isthe main estimated treatment effect, using four and more elections prior to vote-by-mailas a baseline. As in Table 3, the plot clearly captures that turnout increased in the yearimmediately following the introduction of vote-by-mail and turnout was not meaningfullyhigher before the counties adopted voting by mail.
Figure S6 – Vote-by-Mail Reform Modestly Increases Turnout.
-.02
0
.02
.04
.06
With
in-C
ount
y D
iff in
Tur
nout
Rat
e
-3 -2 -1 0Elections Since Treatment
32
S11 Effect of Universal Vote-by-Mail on Republican
Participation
In this section, we show that the non-effects of universal vote-by-mail on Democratic turnoutin Table 2 hold when we instead look at Republican turnout share. The turnout share thatwe construct in columns 1-3 of Table 2 is the number of those who voted in the electionthat are registered as Democrats divided by the total number of those who voted in theelection, regardless of their party affiliation. Because we include third-party and unaffiliatedvoters in the denominator, a non-effect on the Democratic turnout share does not guaranteea non-effect on the Republican turnout share.
In Table S10 we estimate the effects of vote-by-mail on the Republican turnout share.The specifications mirror columns 1-3 in our main results in Table 2. The first column showsthe within-state difference-in-differences estimate, which is a decrease in Republican turnoutshare of approximately two and a half percentage points. The last two columns show thatthis result does not hold once we include county-level trends to control for possible pre-trending if counties that enter vote-by-mail are trending less Republican over time comparedto other counties. In those specifications, the estimate is closer to zero, and in each case wecannot reject the null hypothesis of no effect.
Table S10 – Vote-by-Mail Expansion Does Not Have Large Effectson Republican Share of the Electorate.
Rep Turnout Share [0-1](1) (2) (3)
VBM -0.024 -0.004 -0.007(0.007) (0.004) (0.004)
# Counties 87 87 87# Elections 23 23 23# Obs 986 986 986
County FE Yes Yes YesState by Year FE Yes Yes YesCounty Trends No Linear Quad
Robust standard errors clustered by county in paren-theses.
To investigate the source of pre-trending more, we show the robustness of our difference-in-differences estimate (column 1 of Table S10) based on years included in the sample. Oneconcern with using the voter file data is that voters removed from the voter file over timemay be different from those remaining on the lists. In particular, we know that older votersin 1998 are much less likely to still be in the 2019 voter file we are using. This problemshould be much smaller in elections that were held closer to the time when the voter file wascompiled. In Figure S7, we evaluate the sensitivity of our results to the number of electionsprior to 2018 that we include in the difference-in-differences regression. We find that the
33
estimate attenuates quite a bit when we include only recent elections, which suggests thatregistered Republicans were likely dropping out of the voter file at a higher rate in countiesthat adopted VBM early compared to counties that adopted VBM later.
Overall, even if we take the difference-in-differences estimates in column 1 of Table S10at face value, once we restrict the sample to years where we are more confident in ourestimates of the composition of the electorate, it is clear that we can rule out large effectsof vote-by-mail on the Republican share of the electorate.
Figure S7 – Republican Composition Effects Attenuate as We In-clude Just Recent Years in the Sample.
-.04
-.03
-.02
-.01
0
Effe
ct o
n R
ep S
hare
of E
lect
orat
e
1998 2002 2006 2010 2014First Year Included in Regression
34
S12 Effects On Age of Electorate
In this section, we present evidence on the effect of vote-by-mail on the age composition ofthe electorate. We construct a variable that is the share of the electorate – meaning the shareof those who turn out to vote – that is age 55 or above. We show the effects of vote-by-mailon that outcome in Table S11. The specifications mirror those in columns 1-3 of Table 2. Inour difference-in-differences design in column 1, we estimate that vote-by-mail decreased theshare of the electorate age 55 or above by about one and a half percentage points. Once weinclude county-level time trends in columns 2 and 3, the estimates shrink to be close to zero.In all cases with adjustments for county trends, we cannot reject the null hypothesis thatvote-by-mail does not affect the age composition of the electorate. Though the estimatesare slightly noisier than our main results on the partisan composition of the electorate, weinterpret these results as evidence that vote-by-mail programs to not dramatically changethe age composition of those who turn out.
Table S11 – Vote-by-Mail Expansion Does Not Appear Have LargeEffects on Age Composition of the Electorate.
Turnout Share Age 55+ [0-1](1) (2) (3)
VBM -0.016 -0.004 -0.005(0.012) (0.009) (0.010)
# Counties 87 87 87# Elections 23 23 23# Obs 986 986 986
County FE Yes Yes YesState by Year FE Yes Yes YesCounty Trends No Linear Quad
Robust standard errors clustered by county in paren-theses.
The results in Table S11 rely on the somewhat arbitrary choice of age 55 as the relevantage cutoff. To show that our results are not simply an artifact of this choice of cutoff,in Figure S8 we show our estimates across a range of age cutoff values. For example, theleftmost estimate in Figure S8 shows the estimated effect of vote-by-mail on the share of theelectorate over the age of 30, and we do the same for each value of age from 30 to 65. Thefigure shows the effect of vote-by-mail on the age of the electorate is close to zero across arange of age cutoffs.
35
Figure S8 – Potentially Larger Effect on Electorate Age UsingHigher Age Cutoff.
-.04
-.03
-.02
-.01
0
.01
.02
Effe
ct o
n Sh
are
Ove
r Age
Cut
off
30 35 40 45 50 55 60 65Age Cutoff
36
S13 Effects On Socio-Economic Status and Racial Com-
position of Electorate
In this section, we present evidence on the effect of vote-by-mail on the socio-economic statusand racial composition of the electorate. We construct a variable that captures the share ofthe electorate that lives in census tracts with a 13% or higher poverty rate in the 2011-2016American Community Survey five year sample. Using the same ACS data merged to thevoter file, we measure the share of the electorate that lives in tracts that are more than 70%white.22 We show the effects of vote-by-mail on these two outcomes in Tables S12 and S13.The specifications mirror those in columns 1-3 of Table 2.
Across all specifications in Table S12, the share of respondents coming from high-povertytracts is not affected by vote-by-mail.
Table S13 estimates the effect of universal vote-by-mail on the turnout share from largely-white census tracts – columns 1 through 3 are substantively close to zero and cannot bedistinguished from zero statistically.
Table S12 – Vote-by-Mail Expansion Does Not Appear Have LargeEffects on Socio-Economic Status of the Electorate.
Turnout Share High-Poverty Tracts [0-1](1) (2) (3)
VBM -0.001 0.003 0.002(0.006) (0.005) (0.004)
# Counties 80 80 80# Elections 23 23 23# Obs 904 904 904
County FE Yes Yes YesState by Year FE Yes Yes YesCounty Trends No Linear Quad
Robust standard errors clustered by county in paren-theses.
The results in Tables S12 and S13 rely on the arbitrary choice of the median tract todefine the poverty rate and white-share cutoff. As in Figure S8, to show that our results arenot simply an artifact of this choice of cutoff, in Figures S9 and S10 we show our estimatesacross a range of cutoff values. The figures show the effect of vote-by-mail on the race andpoverty rate of the electorate is close to zero across a range of cutoffs.
2213% is the median poverty tract poverty rate for California and Utah, and 70% is the median tract whiteshare of the population.
37
Table S13 – Vote-by-Mail Expansion Does Not Appear Have LargeEffects on Share of the Electorate from Overwhelmingly-WhiteTracts.
Turnout Share High-White-Share Tracts [0-1](1) (2) (3)
VBM -0.002 0.006 0.004(0.004) (0.004) (0.003)
# Counties 87 87 87# Elections 23 23 23# Obs 986 986 986
County FE Yes Yes YesState by Year FE Yes Yes YesCounty Trends No Linear Quad
Robust standard errors clustered by county in paren-theses.
Figure S9 – No Effect on Electorate Poverty Regardless of Cutoff.
-.01
0
.01
.02
Effe
ct o
n Sh
are
Ove
r Pov
Cut
off
.05 .1 .15 .2 .25 .3Poverty Cutoff
38
Figure S10 – No Effect on Electorate Race Regardless of Cutoff.
-.005
0
.005
.01
.015
Effe
ct o
n Sh
are
Ove
r Cut
off
.3 .4 .5 .6 .7 .8 .9 1White-Share Cutoff
39